Teaching Buildings to Describe Themselves

There is a statistic Integrated Projects likes to cite about its own industry, and it is the kind of statistic that sounds made up until you think about it for a second: eight out of ten buildings do not have accurate as-built plans. The drawing on file, if there is one, describes a building that was intended, not the building that exists. Somewhere between the architect's PDF and the finished structure, a contractor moved a wall, an engineer rerouted a duct, and nobody updated the record.

This is fine, mostly, until someone wants to do something with the building - renovate it, convert it, sell it, comply with a new energy code. Then the gap between the drawing and reality becomes a line item, usually a large and unpleasant one. Renovation budgets rarely blow up on the new work. They blow up on the surprises hidden in the old work.

Integrated Projects, founded in New York in 2018 by Jose Luis Cruz Jr., sells the boring and useful fix. It sends a trained technician with a LiDAR scanner to your property - anywhere in the US, typically within three to five days - and captures a construction-grade 3D scan. Then it converts that scan into standardized BIM and CAD files: the digital, queryable, dimensionally accurate model of what is actually there. You view it in a browser. You measure it, share it, export it.

The clever part is less the laser and more the second step. Turning a point cloud - millions of dots suspended in 3D space - into a usable building model has historically been slow, manual, expert work. Someone traces walls by hand. Integrated Projects points machine learning at the problem, reconstructs the geometry automatically, and keeps humans in the loop for verification. The company says its engine can produce automated floor plans in under an hour.

That sentence, from founder Jose Cruz Jr., is the entire company compressed into one line. It is not a mission statement about disruption or the future of anything. It is a claim that the built world is under-documented, that this costs real money, and that the fix is a data-collection problem worth industrializing.

The results, at least by the company's own numbers, suggest the market agrees. Integrated Projects says it has digitized more than 7,000 buildings and over 52 million square feet across North America, Europe and Australia, serving upward of 1,400 AECO organizations - architecture, engineering, construction and operations. In 2022 alone it digitized more than 1,700 buildings, which it framed as 725% year-over-year growth. It reports a 97% on-time delivery rate, which in a services business is arguably the number that matters most.

None of this is glamorous. It is logistics, scanning hardware, machine-learning pipelines, and quality control, aimed at a problem most people never think about because they assume, wrongly, that the plans already exist. That assumption is the whole opportunity.

To understand why speed is the product and not a feature, consider 25 Water Street in Lower Manhattan - a $500M project that set the record for the largest office-to-residential conversion in the United States, roughly 1,300 apartments carved out of a former office tower. Pavarini McGovern, part of the STO Building Group, needed an accurate existing-conditions model to plan the conversion.

Integrated Projects coordinated multiple scanners at once and delivered an LOD300 existing-conditions BIM in under two weeks. On a project of that scale, the alternative - weeks or months of manual measurement and drafting, with the errors baked in - is not just slower. It is the difference between a schedule that holds and one that doesn't.

The customer list runs through the same logic: institutional landlords, construction managers, architects and operators who own or touch a lot of space and need to know, precisely, what they have. Several of them are also investors, which is either a strong signal or a conflict of interest, depending on how cynical you are feeling. In this case it reads mostly as a signal - the people who own the buildings put money into the company digitizing them.

Integrated Projects sits in a crowded neighborhood - reality-capture and scan-to-BIM includes players like Matterport, OpenSpace and Avvir, plus traditional survey firms and the incumbent Autodesk toolchain. What distinguishes IPX is less any single component than the decision to run the whole loop as a service: capture, convert, verify, deliver, all through one platform, priced per square foot.

The machine-learning side is real and getting more attention. The company employs applied ML scientists - including Thomas Czerniawski, who has spent a decade in computer vision and digital modeling of the built environment - focused on automating the tedious middle of the pipeline. It is a member of NVIDIA's Inception program and works with USD-based 3D formats, aligning with the Omniverse ecosystem for interoperable building data.

If you want the one-sentence version of where proptech is quietly going, it is probably this: skip the flashy visualizations and build the reliable, queryable digital record of every physical building underneath them. That layer is unglamorous, expensive to produce, and genuinely useful. Integrated Projects is building it one scan at a time.